| (00:00) |
SpokenWeb Podcast Intro |
[Instrumental music overlapped with feminine voice]
Can you hear me? I don’t know how much projection to do here. |
| (00:17) |
Hannah McGregor |
What does literature sound like? What stories will we hear if we listen to the archive? Welcome to the Spoken Web podcast, stories about how literature sounds. |
| (00:34) |
Hannah McGregor |
My name is Hannah McGregor, and — |
| (00:36) |
Katherine McLeod |
My name is Katherine McLeod. And each month we’ll be bringing you different stories that explore the intersections of sound, poetry, literature, and history created by scholars, poets, students, and artists from across Canada. |
| (00:49) |
Katherine McLeod |
For many of us who have studied, taught, written, or simply enjoyed poetry, we know that some poets’s work comes alive in performance. I remember a professor in my undergraduate insisting that we read 17th century English poet John Milton’s Paradise Lost aloud since that was how Milton wrote it. He was blind and composed it through dictation. |
| (01:12) |
Katherine McLeod |
In this episode of the SpokenWeb Podcast, Adam Hammond, associate professor of English at the University of Toronto, makes the same argument for T.S. Eliot’s poem “The Wasteland.” Elliot’s poem, he asserts, is written as a collection of voices thrown together, and it’s in the oral performance that these different voices can be heard depending, of course, on the performance decisions of the reader. |
| (01:37) |
Katherine McLeod |
Luckily for us and for Hammond, a lot of people have read “The Wasteland” out loud, including Eliot himself. Even luckier new digital humanities tools, like “Drift” and “Gentle,” now add computational listening into the Modernist Scholars toolkit, allowing us to ask new questions about poetic performances, including the ones that frame this episode. Is Eliot’s reading as dry, monotonous and hopelessly formal as it might sound to a contemporary listener? Or can computational listening help us to hear it a little differently? Here is episode five of season five of the SpokenWeb Podcast: They Do the Police in Different Voices: Computational Analysis of Digitized Performances of T. S. Eliot’s The Waste Land. |
| (02:25) |
Music |
[SpokenWeb theme song begins playing.] |
| (02:35) |
Adam Hammond |
I will never forget the first time I heard T.S. Eliot’s voice, he was reading his poem “The Wasteland.” |
| (02:41) |
Bob Dylan, reading the first four lines of “The Waste Land” for his XM Radio show “Theme Time Radio Hour” |
“April is the cruellest month, breeding / Lilacs out of the dead land, mixing / Memory and desire, stirring / Dull roots with spring rain.” |
| (02:52) |
Adam Hammond |
Okay, I’m just messing with you. That’s not T.S. Eliot. That’s Bob Dylan. This is T.S. Eliot. |
| (02:59) |
T.S. Eliot, reading the first few lines of “The Wasteland” |
“April is the cruellest month, breeding / Lilacs out of the dead land, mixing / Memory and desire, stirring / Dull roots with spring rain.” |
| (03:14) |
Adam Hammond |
I remember the exact thought I had the first time I heard T.S. Eliot’s voice. I thought he was American. I remember the second thought I had as well. I thought he was young. And I remember the third. I thought he was cool. I was 19 in Christian Lloyd’s first year English class at Queen’s University’s International Study Center in Herstmonceux Castle in England. I was supposed to study engineering, but I was able to convince my parents to let me defer my acceptance to Waterloo when I got a scholarship to live in a castle in England for a year and study English. It’s what I wanted to do more than anything. I was there because I loved modernism. I had read Hemingway, Fitzgerald, and Orwell in high school. They all seemed so badass — so free and brave and defiant. |
| (04:06) |
Adam Hammond |
I’d also read T. S. Eliot. I loved “Prufrock.” Even though it spoke about old men with trousers rolled, it seemed like a young person’s poem, about young people’s problems. Like having the courage to be yourself, or rather about not having the courage to be yourself — about “putting on a face to meet the faces that you meet.” I felt like I had taken Eliot’s implicit message in deciding to follow my heart and study English. Prufrock was an unreliable narrator. You were supposed to resist his old-mannish ways; rolled-up trousers were bad. . |
| (04:44) |
Adam Hammond |
But then I was in first year, and we were reading “The Wasteland,” and it seemed infinitely more badass than “Prufrock.” Younger and freer and braver and more defiant. And then my roommate and I found a recording of Eliot reading “The Wasteland,” and he sounded pathetic. He sounded old and lame with a terrible fake British accent. |
| (05:04) |
Adam Hammond |
It was worse than that. He was Prufrock’s dad. It took me years to recover from this. |
| (05:14) |
Adam Hammond |
My name is Adam Hammond. I’m an Associate Professor in the Department of English at the University of Toronto. I never went to Waterloo. I continued to scam my parents all the way through undergrad, saying I would go to law school, that English was the perfect pre-law degree, but that a Master’s was useful prep for all the research lawyers were expected to do. I don’t even remember what excuse I used to justify doing a Ph.D, but I did one. |
| (05:39) |
Adam Hammond |
I wrote my dissertation on three writers, one of whom was T. S. Eliot. I did this because he fit the idea, not because I loved him. I still hadn’t recovered from that experience of hearing his voice. Then, in 2011, right as I was finishing my Ph.D., something wild happened. The venerable printing house Faber and Faber — the very place where Eliot himself worked as poetry editor, serving as modernism’s ultimate gatekeeper — collaborated with an app developer called “TouchPress” to make an iPad app version of “The Wasteland.” |
| (06:12) |
Adam Hammond |
It sounds like this would be a bad thing. But it wasn’t. It was amazing. It completely changed the way I saw the poem. It made it cool again. It had interviews with celebrities, and I mean celebrities had not only heard of “The Wasteland” but they liked it! It had notes you could make disappear. You could swipe right on the words of the poem, and like magic, the thing that’s typescript would appear, scratched to smithereens by Eliot’s pal Ezra Pound. But the real killer feature — the best thing about “The Wasteland” app by far — the clear single reason that I started loving “The Wasteland” again — was that fact that the app included readings of the poem. You touched a line, and you heard it. |
| (06:58) |
Viggo Mortensen, reading “The Wasteland” |
There were a bunch of readings. They were by actual celebrities. One of them was by Viggo Mortensen. One of them was by Alec Guiness. Another was by Fiona Shaw. Jeremy Irons was on there. You could hear the entire Wasteland, all 433 lines of it, read to you by Aragorn, or Obi-Wan Kenobi, or Aunt Petunia Dursley, or Scar! And some of these voices, let me tell you: they were cool. Check this out. |
| (07:29) |
Adam Hammond |
“April is the cruellest month, breeding / Lilacs out of the dead land, mixing / Memory and desire, stirring / Dull roots with spring rain. / Winter kept us warm, covering / Earth in forgetful snow, feeding / A little life with dried tubers.” |
| (07:47) |
Adam Hammond |
That was exactly how I always wanted Eliot to sound. It was my fantasy of the poet’s voice. American, young, and tough. As I put it in an article I wrote for the Toronto Review of Books shortly after the app came out, Mortensen was the anti-Prufrock. It sounded like he was reading the poem from the seat of a Harley Davidson. There were reasons aplenty for nineteen-year-old me to get excited about “The Wasteland.” But there were also reasons for thirty-year-old me, as I then was, to be excited. I had just finished a dissertation about modernism and the phenomenon of dialogism. That was the word that Russian literary critic Mikhail’s Bakhtin used to describe literary texts made up of lots of genuinely competing voices. The characters in a dialogic novel — Bakhtin’s prime example being Dostoevsky — they got into real debates. They disagreed with one another. They disagreed with their author. |
| (08:44) |
Adam Hammond |
The outcome of their debates were totally unpredictable. It was like they were autonomous – independent of their creator. For Bakhtin, dialogic novels were little snow globe versions of healthy democracies — mini public squares. In my dissertation, I argued that modernist dialogism had a political edge: that in an era of rising authoritarianism and mass control, its purpose was to train its readers how to think for themselves, how to cut through all the bullshit, find their own voice in the maelstrom. . |
| (09:17) |
Adam Hammond |
Although Bakhtin thought dialogism existed only in novels, one of my prime examples of a dialogic text was “The Wasteland.” I didn’t really see “The Wasteland” as a poem, you see. I saw it as a kind of novel without a narrator. It was an even more extreme form of dialogism than Dostoevsky. There were voices everywhere! But there weren’t even quotation marks. Everything was a voice, but unlike in a Dostoevskian novel, you couldn’t even say for sure where one voice stopped and where the other began. |
| (09:47) |
Adam Hammond |
When one voice passed the mic to the other voice. This is what I wrote in my review for The Toronto Review of Books — the part where I was explaining why I was so excited about all the audio readings in “The Wasteland” app: “The focus on oral performance [in the app] works especially well with The Waste Land, because it is a poem that demands so emphatically to be read aloud—and indeed only really makes sense once you begin to consider it in the light of oral performance. As Eliot’s original title for the poem, He Do the Police in Different Voices, reminds us, the basic unit of The Waste Land is the voice. But though the poem is built from multiple distinct voices, it does not tell us where they begin or end, or what each is like, nor does it provide a dramatis personae or indicate its speakers. These voices thus only really become apparent in oral performance, where the reader must decide on their cast of characters, and give each one a recognizable personality. Every reading of the poem is thus an interpretation of some of its most fundamental questions.” |
| (10:56) |
Alex Guinness, reading “The Wasteland” |
Yes, that’s right, Eliot’s working title for “The Wasteland” was “He Do the Police in Different Voices” — a reference to something someone says about a character in Charles Dickens’s “Our Mutual Friend,” who animates his reading of newspaper stories by giving the police men funny voices. And this is all more evidence — because it’s a very cool title — of Eliot’s fundamental and latent badassness. I’ll show you what I mean about the voices in the poem. This is Alex Guiness, aka Obi-Wan Kenobi, reading the opening of the poem. Listen carefully and you’ll hear him switching into the voice of “Marie.” |
| (11:35) |
Adam Hammond |
“April is the cruellest month, breeding Lilacs out of the dead land, mixing Memory and desire, stirring Dull roots with spring rain. Winter kept us warm, covering Earth in forgetful snow, feeding A little life with dried tubers. Summer surprised us, coming over the Starnbergersee With a shower of rain; we stopped in the colonnade, And went on in sunlight, into the Hofgarten, And drank coffee, and talked for an hour. Bin gar keine Russin, stamm’ aus Litauen, echt deutsch. And when we were children, staying at the archduke’s, My cousin’s, he took me out on a sled, And I was frightened. He said, Marie, Marie, hold on tight. And down we went. In the mountains, there you feel free. I read, much of the night, and go south in the winter.” |
| (12:35) |
Fiona Shaw, reading “The Wasteland” |
Now listen to Fiona Shaw — aka Aunt Petunia Dursley — read the same part. She does the voice even more clearly. |
| (12:44) |
Adam Hammond |
“April is the cruellest month, breeding Lilacs out of the dead land, mixing Memory and desire, stirring Dull roots with spring rain. Winter kept us warm, covering Earth in forgetful snow, feeding A little life with dried tubers. Summer surprised us, coming over the Starnbergersee With a shower of rain; we stopped in the colonnade, And went on in sunlight, into the Hofgarten, And drank coffee, and talked for an hour. Bin gar keine Russin, stamm’ aus Litauen, echt deutsch. And when we were children, staying at the archduke’s, My cousin’s, he took me out on a sled, And I was frightened. He said, Marie, Marie, hold on tight. And down we went. In the mountains, there you feel free. I read, much of the night, and go south in the winter.” |
| (13:38) |
Viggo Mortensen |
Viggo Mortensen sure sounds cool, and I can’t help it if he’s still my favourite reader of the poem, but it’s pretty hard to tell if he’s doing a voice there. |
| (13:45) |
Adam Hammond |
“April is the cruellest month, breeding Lilacs out of the dead land, mixing Memory and desire, stirring Dull roots with spring rain. Winter kept us warm, covering Earth in forgetful snow, feeding A little life with dried tubers. Summer surprised us, coming over the Starnbergersee With a shower of rain; we stopped in the colonnade, And went on in sunlight, into the Hofgarten, And drank coffee, and talked for an hour. Bin gar keine Russin, stamm’ aus Litauen, echt deutsch. And when we were children, staying at the archduke’s, My cousin’s, he took me out on a sled, And I was frightened. He said, Marie, Marie, hold on tight. And down we went. In the mountains, there you feel free. I read, much of the night, and go south in the winter.” |
| (14:34) |
T.S. Eliot, reading “The Wasteland” |
But hold on. Eliot wrote “The Wasteland.” He’s the one who called it “He Do the Police in Different Voices.” He knew all about the voices. So does he do them? For better or worse, that exact same recording that I heard way back when I was nineteen and living in a castle in southern England — it was on the app, too. So, does Eliot do Marie’s voice in the opening of the poem? I hate to do it to you, but let’s listen to it again. |
| (15:04) |
Adam Hammond |
“April is the cruellest month, breeding Lilacs out of the dead land, mixing Memory and desire, stirring Dull roots with spring rain. Winter kept us warm, covering Earth in forgetful snow, feeding A little life with dried tubers. Summer surprised us, coming over the Starnbergersee With a shower of rain; we stopped in the colonnade, And went on in sunlight, into the Hofgarten, And drank coffee, and talked for an hour. Bin gar keine Russin, stamm’ aus Litauen, echt deutsch. And when we were children, staying at the archduke’s, My cousin’s, he took me out on a sled, And I was frightened. He said, Marie, Marie, hold on tight. And down we went. In the mountains, there you feel free. I read, much of the night, and go south in the winter.” |
| (15:59) |
Adam Hammond |
My answer is: I don’t know, sorta. Like he wants to, but he’s too shy. Too bad at performing, too much of a poet, not enough of an actor. |
| (16:12) |
Adam Hammond |
Around the time the app came out, I finished my PhD and became obsessed with the so-called “Digital Humanities.” I started a couple of digital projects for exploring multi-voicedness in “The Wasteland” around this time. I made a website called “He Do the Police in Different Voices,” of course, that presented the poem as a play. Working with a big class of undergraduates, we decided on one way of dividing the poem up into characters and then we created a digital edition with names and special fonts for all the different voices. I also did a computational text analysis project with two computational linguists, Julian Brooke and Graeme Hirst. We used a variety of natural language processing techniques to see where a computer might detect voice switches in the poem. All those moments where the computer thought the mic was being handed from one character to another. |
| (17:01) |
Alec Guiness |
The results of this were really interesting. Approaching the poem with the mind of a machine, the algorithm we developed found switches in places I hadn’t ever imagined them. And on reflection, a lot of these seemed really on point. For instance, I had always heard a switch at “winter kept us warm,” but the computer didn’t see one there. It thought there was a switch at “summer surprised us,” where I personally had never seen a switch. But then listening back to Alec Guiness, that’s exactly where he goes into the voice of “Marie.” |
| (17:33) |
Adam Hammond |
“Summer surprised us, coming over the Starnbergersee / With a shower of rain;” |
| (17:39) |
Adam Hammond |
The algorithm had opinions and I found these opinions worthwhile. |
| (17:45) |
Adam Hammond |
In a 2005 article on computational analysis of literature, Julia Flanders wrote that we shouldn’t look at computers for objective answers about literary interpretation. As she put it, computers shouldn’t be seen as “factual substantiator whose observations are different in kind from our own — because more trustworthy and objective — but rather computrers should be seen as a device that extends the range of our perceptions to phenomena too minutely disseminated for our ordinary reading.”. |
| (18:18) |
Adam Hammond |
They’re not there to confirm our subjective readings of a poem as objectively true. They’re there to challenge our readings with their own readings, which are no more objective than ours, but are definitely different based on things we humans don’t even notice when reading. What I couldn’t do at that time was analyze the audio recordings from the app. I could only look at text, but I was definitely curious about analyzing these audio recordings. I had my own human feelings about which readings on the app were the most dynamic or the most polyvocal or just the coolest. In other words, I had feelings about which reader did the police in different voices better than the others. What I didn’t have was any computational voice to bounce these ideas off of. |
| (19:10) |
Adam Hammond |
In the decade that followed the tools that I dreamed of were developed, many of them by teams led by a poetry scholar named Marit MacArthur. Working with scientists, programmers, humanists and students, she led the development of a set of computational and theoretical tools to analyze audio recordings of performances of poetry. MacArthur’s method works from only two data points: pitch and timing. To get the timing information, we used a program called “Gentle,” designed in collaboration with MacArthur. Basically, we fed the program all of our recordings of all of the app’s performances of “The Wasteland,” and it told us exactly when each word in the poem was spoken and how long the gaps were between these words. My research assistant Jonathan Dick, manually corrected the output of each of these, which was a huge job. This timing data allowed us to calculate how quickly each reader reads in words per minute. Are they fast or are they slow? It calculates the average length of their pauses. So how long do they wait between words? It tells us how often they pause, and it also tells us what kind of rhythms their pauses create. Are they monotonous or do they change like, like, that. |
| (20:40) |
Adam Hammond |
Like a William Shatner kind of a complexity of pauses. We used another program called “Drift,” also designed in collaboration with MacArthur, to get pitch information. “Drift” divides the recording into segments of 100th of a second long and gives the fundamental frequency in Hertz for each of these segments. This data can tell us the pitch range in octaves of a given performance. So this just tells you how [Adam deepens his voice] low does the reader go and how [Adam hightens his voice] high. It gives you the pitch speed also in octaves, so this would be like [Adam eximplifies the pitch speef]: if you go from low to high steadily, there’s a speed. And then pitch acceleration, which is like [Adam mimics an engine-like sound] when the pitch changes in these kind of quickly accelerating fashions. All of these can be used as measures of like how dynamic or dramatic a performance is. In a path-breaking 2018 article, “Beyond Poet Voice” in the journal Cultural Analytics, MacArthur and her collaborators Georgia Zellou and Lee M. Miller proposed four dimensions of poetic performance, and argued that these can be described quantitatively using only this timing and pitch data. |
| (22:01) |
Adam Hammond |
So one dimension they called “formal,” that’s readings with predictable rhythms and slower speech. So that’s all from the timing data. They also had a dimension called “conversational.” This is someone who reads with less predictable rhythms and faster speech. So again, you can get all of this from the timing data. They had a dimension called “expressive.” This is someone with a wide pitch range and highly contrasting pitch – up and down, high speed, high acceleration. The final dimension they called “dramatic,” which features long unpredictable pauses, again, timing related. With these tools and theories in place, we were equipped to dig into exciting research questions about the performances on “The Wasteland” app. And to get answers to these questions, both from human readers and machine readers. Our broad questions were: number one, where do readers of “The Wasteland” do voices? Where do these voice switches occur? Where does one voice pass the mic to the other? Number two: how do readers of “The Wasteland” do voices? What aspects of timing and pitch do they alter to indicate voice switches? Number three, this is a big question: Is dialogism or multi-voice a property of texts or performances? Is it inherent in the text or is it something that is only brought out, even created, in performance? |
| (23:34) |
Adam Hammond |
For the digital tools in particular, we had two questions. Number one: can analysis of pitch and timing information capture the way that different readers do voices or is there something other than just pitch and timing that you need to really understand this? Can this data tell us more about the way that readers do these voices than regular human listening can? Can the computational analysis reveal features that humans, more specifically literary scholars who know “The Wasteland” really, really well, can’t notice? Can they, for instance, shake my long held belief that T.S. Eliot is a terrible reader of his own poem? Here’s what we did. Me and my research assistant, Jonathan Dick, each listened to every reading on “The Wasteland” app and wrote up detailed answers to a series of questions about our subjective impressions. Number one: using subjective criteria, place the reader along the dimensions, formal, conversational, expressive, dramatic, including hybrids of these that are identified in the article beyond poet voice. |
| (24:37) |
Fiona Shaw, reading “The Wasteland” |
Number two: come up with two to three moments or passages that you feel best exemplify the above analysis. Number three: note any cases where the four dimensions or the explanation of these dimensions seem inadequate. Number four: briefly described how well, and just “how,” each reader does the voices in the poem. Number five: pick out a couple of passages where the reader clearly does a voice or conversely remarkably fails to do a voice. So that was our questionnaire. I’ll give you a couple of examples we agreed on using a passage from the poem that really brought out the differences. For instance, we both thought that Fiona Shaw was an expressive, dramatic reader, using a wide pitch range, highly contrastive pitch and incorporating a bunch of dramatic pauses. Have a listen, |
| (25:28) |
Adam Hammond |
“But at my back from time to time I hear The sound of horns and motors, which shall bring Sweeney to Mrs. Porter in the spring. O the moon shone bright on Mrs. Porter And on her daughter They wash their feet in soda water Et O ces voix d’enfants, chantant dans la coupole! Twit twit twit Jug jug jug jug jug jug So rudely forc’d.” |
| (26:05) |
Adam Hammond |
We also both thought that Viggo Mortensen was a formal inexpressive reader. He speaks slowly. His rhythms are steady and predictable, and he doesn’t do much with pitch. |
| (26:16) |
Viggo Mortensen, readibg “The Wasteland” |
“But at my back from time to time I hear The sound of horns and motors, which shall bring Sweeney to Mrs. Porter in the spring. O the moon shone bright on Mrs. Porter And on her daughter They wash their feet in soda water Et O ces voix d’enfants, chantant dans la coupole! Twit twit twit Jug jug jug jug jug jug So rudely forc’d. / Tereu” |
| (26:46) |
Adama Hammond |
By the way, as a kind of control, we also had the old MacOS text-to-speech voice, “Fred,” read the poem, and we analyzed his reading, too. Subjectively, we called it “formal-inexpressive,” just the same as Viggo. |
| (27:02) |
MacOS “Fred,” reading “The Wasteland” |
[Audio recording of MacOS text-to-speech voice, “Fred,” reading “The Wasteland” in a robotic and monotonous tone.] |
| (27:16) |
Adam Hammond |
Then it was time to run the numbers and see how it all shook out computationally. Jonathan and I, and the computer, in many ways weren’t far off in our interpretations of the performances. The computer agreed that Shaw was expressive-dramatic, hardly a surprise. But the computer thought Viggo’s rhythms were a bit more varied that we’d given him credit for: the computer agreed he was inexpressive but, based on his timing data, called him conversational rather than formal. Notably, the computer saw Fred the same way, as conversational and inexpressive. But there was one reader where Jonathan and I just couldn’t agree: Eliot. Have a listen to Eliot reading that same passage. Where would you place him? |
| (28:06) |
T.S. Eliot, reading “The Wasteland” |
“But at my back from time to time I hear The sound of horns and motors, which shall bring Sweeney to Mrs. Porter in the spring. O the moon shone bright on Mrs. Porter And on her daughter They wash their feet in soda water Et O ces voix d’enfants, chantant dans la coupole! Twit twit twit Jug jug jug jug jug jug So rudely forc’d. Tereu” |
| (28:43) |
Adam Hammond |
Jonathan thought that Eliot was formal and inexpressive. In his notes he said, “Eliot is a formal speaker. His tone is neutral and slow. He is also less expressive since his pitch range is rather narrow and non-contrasting. Indeed one might describe him as monotonous.” Jonathan’s impressions agree with what a lot of critics have thought about Eliot over the years. For example, Jason Camlot, in his 2019 book “Phonopoetics” speaks of Eliot’s “calculatedly numb or mechanical delivery”; his delivery, Camlot says, “is robotically liturgical, or […] mechanically oracular”. Now that’s exactly how 19-year-old me felt about Eliot’s reading, and that’s exactly what I hated about it. And yet when I listened to Eliot again to put my subjective responses together, I couldn’t help but disagree with my younger self. In my notes I wrote, “He’s all over the map. There is a lot of formal, but I get the sense that this happens when he’s in ‘formal’ voices. I almost always have the sense that he’s trying to ‘do’ a voice. He is conversational in several voices. One failing he seems to have is varying his rhythms, which are generally monotonous. Definitely, he is expressive in parts, but I get the sense that he is more trying than succeeding.” |
| (29:47) |
Adam Hammond |
Now, I wasn’t alone in hearing voices in Eliot’s reading of “The Wasteland,” or at least the attempt to do these voices. For instance, to go back a century, when Eliot read “The Wasteland” to his friend Virginia Woolf at her house in 1922, she wrote up some subjective impressions of her own in her diary. She wrote, “Eliot dined last Sunday & read his poem. He sang it & chanted it rhythmed it. It has great beauty & force of phrase: symmetry; & tensity.” “One was left,” Woolf said, “with some strong emotion.” Virginia Woolf seems to have considered Eliot an expressive-dramatic reader. |
| (30:51) |
Adam Hammond |
Alas, the computer disagreed with Virginia and I. Even Mortensen and Fred got to be formal and conversational. Eliot’s reading was the only one the computer saw as formal and inexpressive. But maybe I just wasn’t running the numbers right. For all of the results I’ve talked about so far, the computer was giving us results for the whole performance. On average, it would look at timing and pitch data for the full poem, all 20 to 30 minutes of it. It’s a long poem. And then give us data like average pause, length, rhythmic complexity of pauses on average, average pitch acceleration for the whole 20 to 30 minute performance. But maybe that wasn’t what was most interesting or useful in terms of calculating the numbers because the poem is made up of lots of different voices after all. What did I care about average numbers for the whole poem? |
| (31:53) |
Adam Hammond |
That would only make sense if there was only one voice for the whole poem. It would be like putting the Norton Anthology in a text analysis algorithm and getting it to tell me what the average style of a hundred different writers was like. Useless, right? You wanna look at the style for each of the individual writers. Now, what we needed to do was compare the way that the different readers did particular voices. How are they reading here and how does that compare to how they’re reading over here? Do they vary the voice from passage to passage? So for the next stage of our analysis, we identified three passages in the poem that are clearly in different voices. We started with a very formal conventionally poetic passage. We call it the “burnished throne “passage. Then a very informal passage, the famous bar scene, and then a passage made up of a wide variety of voices all stuck together, the “Madame Sosostris” passage. Now we would expect a good reader, a reader who really does the voices, to make a huge contrast between the “burnished throne” voice and the “bar scene” voice. If they really get the poem, they’ll do everything they can to make these voices sound different from one another. |
| (33:14) |
Adam Hammond |
Well, can you believe it? Analyzing all the performances of these three passages with our pitch and timing tools and then comparing the numbers between passages, Eliot is actually the one who varies his reading the most. In terms of words per minute, pitch speed, pause rate, average pitch, his readings are right up there as the most contrasting. Whereas someone like Fiona Shaw is varying her voice all the time all over the place, Eliot is the one, or one of the ones, who varies his voice the most from passage to passage. So have a listen for yourself. Here is Eliot reading the conventionally poetic “burnished throne” passage. |
| (34:01) |
T.S. Eliot, reading “The Wasteland” |
“The Chair she sat in, like a burnished throne, Glowed on the marble, where the glass Held up by standards wrought with fruited vines From which a golden Cupidon peeped out (Another hid his eyes behind his wing)” |
| (34:19) |
Adam Hammond |
And here he is reading the colloquial “bar scene.” |
| (34:23) |
T.S. Eliot, reading “The Wasteland” |
“If you don’t like it you can get on with it, I said. Others can pick and choose if you can’t. But if Albert makes off, it won’t be for lack of telling. You ought to be ashamed, I said, to look so antique. (And her only thirty-one.) I can’t help it, she said, pulling a long face, It’s them pills I took, to bring it off, she said.” |
| (34:42) |
Adam Hammond |
Okay, he’s not the best actor, but you see he’s really trying to do a cockney accent in the bar scene. The computer also placed Elliot among the most dynamic readers for the Madame Sosostris passage. That’s one where there are a lot of voices and we expect a lot of internal variation. This is one where the overall numbers for the passage might actually be interesting. Indeed, although the computer saw Eliot’s performance as overall formal and inexpressive, it actually interpreted his performance of this passage as dramatic and expressive. Let’s listen first to Alex Guinness, aka Obi-Wan Kenobi, reading this passage and you can really hear the way he does the voices. There’s a difference between a kind of a neutral and narrator like voice. A prophetic voice that speaks the lines “Those are pearls that were his eyes!”, and the Eastern European-accented voice of Madame Sosostris herself. |
| (35:39) |
Alex Guinness, reading “The Wasteland” |
“Madame Sosostris, famous clairvoyante, Had a bad cold, nevertheless Is known to be the wisest woman in Europe, With a wicked pack of cards. Here, said she, Is your card, the drowned Phoenician Sailor, (Those are pearls that were his eyes. Look!) Here is Belladonna, the Lady of the Rocks, The lady of situations. Here is the man with three staves, and here the Wheel, And here is the one-eyed merchant, and this card, Which is blank, is something he carries on his back, Which I am forbidden to see. I do not find The Hanged Man. Fear death by water. I see crowds of people, walking round in a ring. Thank you. If you see dear Mrs. Equitone, Tell her I bring the horoscope myself: One must be so careful these days.” |
| (36:39) |
Adam Hammond |
The computer saw Guinness as the most dramatic and expressive of all readers in this scene. But guess who it saw as the second most dramatic and expressive? Our friend T.S. Eliot. And really, if you listen, you can see why. Eliot does all the same voices as Guinness. He does that narrator at the start. He does the prophetic voice for the “pearl’s eyes” line, and he does Madame Sosostriswith that same accent. |
| (37:06) |
T.S. Eliot, reading “The Wasteland” |
“Madame Sosostris, famous clairvoyante, Had a bad cold, nevertheless Is known to be the wisest woman in Europe, With a wicked pack of cards. Here, said she, Is your card, the drowned Phoenician Sailor, (Those are pearls that were his eyes. Look!) Here is Belladonna, the Lady of the Rocks, The lady of situations. Here is the man with three staves, and here the Wheel, And here is the one-eyed merchant, and this card, Which is blank, is something he carries on his back, Which I am forbidden to see. I do not find The Hanged Man. Fear death by water. I see crowds of people, walking round in a ring. Thank you. If you see dear Mrs. Equitone, Tell her I bring the horoscope myself: One must be so careful these days.” |
| (38:09) |
Adam Hammond |
I think I may hear another voice right at the end there, a Cockney man that Guinness doesn’t do. |
| (38:16) |
Adam Hammond |
So where does all this leave us? Well, for one thing, thanks to MacArthur and our collaborators, we now have accessible and powerful computational tools for analyzing poetic performances. The computational analysis of pitch and timing data, permitted by tools like “Gentle” and “Drift,” produce results that correspond well to human listeners’s subjective impressions. In other words, to tools work, which is important. These subjective impressions vary between individual listeners or even between the same listener over time. When I heard Eliot when I was 19, all I heard was a fake English accent and the worst example of a monotonous formal poet’s voice. Jonathan listening today heard something similar, but the me of today disagreed with both, hearing a genuine attempt, however clumsy, however awkward, to really do all the different voices in the poem. |
| (39:09) |
Adam Hammond |
And I mean, he was a poet, after all. He was not a famous actor. When they were casting “Star Wars” and the “Lord of the Rings” and “Harry Potter,” no one was knocking on Eliot’s door. Okay, he was long dead by then, but you get my point. Maybe we need to cut Eliot a little bit of slack. |
| (39:28) |
Adam Hammond |
I think this whole experiment shows why some of us are so drawn to computational analysis of literature and literary performance. Whether you’re working with text or audio, computational tools provide different ways of attending to the work of art, different ways of listening and reading. Computers just notice things that humans don’t, and sometimes those differences can be really interesting. They give us another voice, another perspective to bounce our ideas off of. I still don’t think Elliot was a good reader of his own work, but I do think that he was trying to be a good reader of his own work and the computer seems to agree with me. |
| (40:11) |
Adam Hammond |
Thanks so much for listening, and if you’re interested in this kind of stuff, please feel free to draw me a line. |
| (40:16) |
Music |
[Electronic musis begins playing.] |
| (40:28) |
Katherine McLeod |
The SpokenWeb Podcast is a monthly podcast produced by the Spoken web team as part of distributing the audio collected from and created using Canadian literary archival recordings found at universities across Canada. |
| (40:40) |
Hannah McGregor |
This month’s episode was produced by Adam Hammond. The Spoken Web podcasting team is: supervising producer Maya Harris, sound designer James Healy, transcriber Yara Ajeeb, and co-hosts Katherine McLeod and me, Hannah McGregor. |
| (40:58) |
Katherine McLeod |
[Spokenweb Podcast outro music begins playing] To find out more about SpokenWeb, visit spokenweb.ca and subscribe to SpokenWeb Podcast on Apple Podcasts, Spotify, or wherever you may listen. If you love us, let us know. Rate us and leave a comment on Apple Podcasts, or say “hi” on our social media at Spoken Web Canada. Stay tuned to your podcast feed later this month for shortcuts with me, Katherine McLeod, short stories about how literature sounds. |
